Drowsy Driver Monitoring Using Machine Learning and Visible Actions

Abstract

Driver sleepiness has become a leading cause of traffic accidents and fatalities in recent years. The goal of this research is to find a way to identify driver fatigue and provide early warning so that people can be saved. Using image processing techniques, a camera captures video of the driver's face and measures the status of their eye and mouth opening ratios and delivers a warning if necessary. This is a real-time system. There are a variety of methods for determining whether a driver is drowsy, but this one is absolutely non-intrusive and has no effect on the driving in any way. The per-closure value of the eye is taken into account for the identification of drowsiness. Consequently, the driver is classified as sleepy if the closing of the eye exceeds a predetermined threshold. Offline testing of various machine learning algorithms has also been conducted. Support Vector Machine-based classification has a sensibility of 95.58 percent and a specificity of 100 percent.